GraphTalks Rome - The Italian Business Graph
- 1. Rome, 8th March 2017
Fueling innovation in finance
The Italian business graph
Stefano Gatti – Head of Innovation & Data Sources
- 4. 4
Our business & figures
CREDIT INFORMATION
MARKETING SOLUTIONS
CREDIT MANAGEMENT
1000 report/min
✓ Documents
50 million
✓ Lines of code
34,000
✓ Customers
59 million
✓ Payments data
1,800
✓ People
377 million Eur (2016)
✓ Revenue
Protect your business against credit risk
Find new business opportunities
Optimize credit management processes
- 5. 5
Our “big data”
Web Data
Open Data
Proprietary data
Official data
Chamber of
Commerce
official data
A
c
c
u
r
a
c
y
C
o
m
p
l
e
x
i
t
y
- 7. 7
The beneficial owner of a company
Level 1: 10% Level 3: 24.4% Level 4: 34.2%
ACME spa
Soc. A (40%)
Soc. B (50%)
Willy (10%)
Soc. A1 (40%)
Soc. A2 (60%)
Soc. B1 (40%)
Soc. B2 (60%)
Willy
(40%)
Duffy
(60%)
Willy
(40%)
Soc. B1.1 (60%)
Soc. B2.1 (50%)
Speedy
(50%)
Willy
(90%)
Bunny
(10%)
Level 2: 10%
- 8. 8
2012: a real-time & IT-cost problem … solved!
Relational database Neo4J
Average time 369 ms (till 4th level)
> 12 sec (from 4th level)
39 ms (till 15th level)
• Computed on 2.238.000 companies &
2.756.000 distinct people
• About 150.000 computation/hour
• More than 350.000 push notification to
our customers (mainly banks) in last 6
months
Neo4J in cluster configuration
(two standard servers 12GB
RAM) with usage of Traversal
Framework & In-Memory
Caching. Average usage of CPU
< 5%
Level Beneficial Owners
1 3.379.755
2 202.194
3 44.589
4 11.236
5 3.403
6 1.364
7 464
8 304
9 123
10 61
11 36
12 16
13 6
14 4
15 2
2016 - Statistics
- 9. 9
Aracne Administration Dashboard
5 kinds of relationships :
• Actual Shareholders
• Past Shareholders
• Actual Managers
• Past Managers
• M&A Operations
2 kinds of nodes :
• Companies
• People (Managers & Shareholders)
- 11. 11
2015 - Graph4you: from datascientists to the market
Born as a living datascientist lab for exploring data has easily matched some use-cases of our customers
graph4you.cerved.com
http://developer.cerved.com/
Nodes
• Companies
• Registered 8,1 M
(of which) not active 1,9 M
• Not Registered 1,6 M
• Not Italian 320 K
• People 16,4 M
• Locations 4,5 M
• Customer data
Relationships
• Manager 15,6 M
• Shareholders 8 M
• Building property 14,7 M
• Current address 16,7 M
• Beneficial Ownership 3,6 M
• M&A 100K
Filter:
• Kind of nodes
• Connection degree
• Kind of relationship
• Property of relationship
=͂ 30 M
Nodes
=͂ 59 M
Egdes
- 12. 12
Graph4you
Architecture diagram
WEB APP
FRONT - END
SPRING BOOT
LINKURIOUS.JS
BOOTSTRAP
CUSTOM CODE
JASPER REPORT
SPRING BOOT
SPRING
JAVA
Shortest Path
Traversal
REST API
BACK - END
API
Transactional
HTTP API
INGESTION
Graph fileMySQL
- 14. 15
Graph4You - API
User AppUX
Developer
API Dev Portal
API Graph4You
Cerved
“big data”
Consumer Solution
Cerved Smart
Data API
Framework
- 17. 18
Cerved corporate linkages
Made using Neo4J as OLTP database and computing in-memory the proprietary algorithm
Made in 2014 in partnership with SPAZIODATI
22%
Better recall
30%
More precision
with more data &
managing cycles
Near
Real Time
Levels of depth Frequency
1 178.542
2 36.435
3 5.044
4 913
5 137
6 44
7 13
8 1
9 2
Tot 221.131
Analytics on Neo4J
- 20. 21
Cerved future in graph journey …
Critical Chain (Ex: Contagion Risk)
next
Corporate linkages in Graph4You
Work in Progress
Company similarity
Work in Progress
- 21. 22
Cerved big data journey: «lessons learned»
Augmented data
Extracting value from existing
data with technology
Extracting new data with
technology
- 22. 23
Future is not what is used to be …
"The illiterate of the 21st century will not be those
who cannot read and write, but those who cannot
learn, unlearn, and relearn."
Alvin Toffler